Skip to content

Spatial Analysis

Goal: Move beyond simple geoprocessing into pattern detection — find hot spots, compute density, model suitability, and measure accessibility.

What you'll learn

  • Hot spot analysis (Getis-Ord Gi*)
  • Kernel density and heatmaps
  • Site suitability modeling
  • Accessibility and service-area analysis

Spatial autocorrelation

Tobler's Law again: near things are more related than distant things. Spatial analysis quantifies that.

Statistic What it tells you
Moran's I Is the dataset clustered, dispersed, or random?
Getis-Ord Gi* Where are the hot spots and cold spots?

Moran's I

A single number for the whole dataset:

  • ≈ +1 → strong clustering (similar values near each other)
  • ≈ 0 → random
  • ≈ -1 → dispersed (chess-board pattern)

Tool: Spatial Autocorrelation (Moran's I)

Hot Spot Analysis (Getis-Ord Gi*)

Adds a value to each feature: is this a statistically significant hot or cold spot?

Tool: Hot Spot Analysis (Getis-Ord Gi*)

Use case
Crime hot spots in a city
Disease clusters
Areas of high foreclosure rate
Sales territories under-performing

Don't confuse hotspot and density

  • Density (heatmap): "lots of points here"
  • Hotspot (Gi*): "lots of points and the surrounding area also has lots — statistically unusual"

Density analysis

Kernel density (heatmap)

Converts a point layer into a continuous raster surface. Each point spreads its "weight" over a search radius (kernel).

Use case
Crime heatmap
Animal sightings density
Traffic accident hotspots
Population density from points

Tool: Kernel Density (Spatial Analyst).

Point density vs Kernel density

  • Point Density — count of points per cell area
  • Kernel Density — smoothed surface using a quadratic kernel; usually preferred

Suitability analysis

Find the best location by combining multiple criteria.

flowchart TD
    Goal[Find best site<br/>for a new park] --> C1[Within 1 km<br/>of population center]
    Goal --> C2[Slope < 10%]
    Goal --> C3[Not in floodplain]
    Goal --> C4[Land = vacant]

    C1 --> Combine[Weighted Overlay]
    C2 --> Combine
    C3 --> Combine
    C4 --> Combine
    Combine --> Result[Suitability raster<br/>0 = bad, 100 = ideal]

    classDef goal fill:#4338ca,stroke:#312e81,color:#fff
    class Goal goal
    classDef cri fill:#eef2ff,stroke:#4338ca,color:#312e81
    class C1,C2,C3,C4 cri
    classDef tool fill:#fef3c7,stroke:#f59e0b,color:#92400e
    class Combine tool
    classDef out fill:#dcfce7,stroke:#10b981,color:#065f46
    class Result out

Steps (raster approach)

  1. Convert each criterion into a raster.
  2. Reclassify each raster to a common scale (1–10 or 0–100).
  3. Weight the rasters (some criteria matter more).
  4. Run Weighted Overlay or Weighted Sum.
  5. Identify cells with the highest score.

→ Project: Site Suitability Analysis

Accessibility analysis

Service area

"Within a 10-minute drive of this hospital, who lives there?"

Tool: Service Area (Network Analyst). Requires a road network dataset.

Origin-destination (OD) cost matrix

"What's the travel time from each block to its nearest grocery store?"

Tool: OD Cost Matrix (Network Analyst).

Closest facility

"If a 911 call comes in at this point, which 3 fire stations are closest by drive time?"

Tool: Closest Facility (Network Analyst).

→ See Network Analysis.

Spatial statistics quick reference

  • Average Nearest Neighbor

    Are your points clustered or dispersed compared to random?

  • Standard Deviational Ellipse

    Direction & spread of a point pattern.

  • Mean Center / Median Center

    "Where is the center of gravity of these points?"

  • Cluster and Outlier Analysis (Anselin Local Moran's I)

    Identify high-high, low-low, and outlier features.

Time-aware analysis

Adding time to spatial analysis unlocks change detection.

  • Space-Time Cube (ArcGIS Pro) — analyze patterns over space and time together.
  • Emerging Hot Spot Analysis — find new, intensifying, persistent, or sporadic hot spots.

Analysis ≠ Map

Analysis-first thinking

Decide what question you're answering before opening ArcGIS Pro. The hardest part of spatial analysis is choosing the right tool — not running it.

Write the question on paper:

"Are crime incidents in 2024 statistically clustered near transit stops?"

Now you know which tools to reach for: a hot spot analysis on crimes, a buffer on transit, and a comparison.

Practice

Mini analysis project

Pick a city. Download:

  1. Crime points (any open-data portal)
  2. Transit stops
  3. Income by census tract

Then:

  1. Run Hot Spot Analysis on crime counts per tract.
  2. Run Kernel Density on crime points.
  3. Compute the average distance from each tract centroid to the nearest transit stop.
  4. Correlate hot spots with income and transit access.
  5. Write a 1-page summary with a single map.

That's a portfolio project.

Next up

→ Web GIS — sharing your maps online.